Univariate and multivariate time series models and discriminant techniques are being applied to various data bases consisting of short series of measurements of serum biochemistries in healthy subjects and patients with myocardial infarction. The purpose is to gain practical experience in the use of these statistical predictive techniques to detect changes and trends within individuals, taking into account biological variation and measurement error. The time scale of these series varies from daily to weekly, 6-month, and 12-month intervals between observations. Parallel computer-based simulation studies are also underway, particularly to estimate the relative sensitivities and specificities of multivariate and univariate forecasting methods. Mathematical investigations into the properties of a new stochastic model of linear change are continuing.